Performance Testing

[cyclebrain]

After listening to my completed EZ DAC for about a week now I decided
that it was time for me to back up my subjective listening with some
measured data.
I noticed that the output level was quite low relative to my other
input sources.
This was verified when I played a 1KHz 0dB test signal and measured
380mv output into a 47k ohm load. The redbook standard for CD
specifies a 2 Vrms output.
A quick check of the circuit, 7.8ma from the 1794 at 0dB into a 25 ohm
I/V converter and then fed into a gain stage of 6.67 (1000ohms/
150ohms) gives similar calculations for output level.
At this point I decided stop my testing and bring the output level up
to spec. Since noise is often a fixed level, having a low signal level
might compromise S/N ratio.
I decided to change to I/V resistor to 50 ohms without any real clue
as to what kind of loads the 1794 could drive. It should double the
input voltage to the amplifier stage though.
I then replaced the 1K feedback resistors with 2.7K resistors.
While making these changes I got the -12V regulator IC caught in one
of the grooves in the side of my case and broke two leads off. No more
testing until I can replace the part.
DOH!

[ezkcdude]

For a 25R I/V resistor followed by a gain of let's say 10X, and
PCM1794 of roughly 8mA, isn't the output 25*8*10=2V? Am I doing this
calculation wrong? I got it from a reliable source. With the DAC I
built, I thought the output was pretty close to my Squeezebox which is
around 2V I think.
-evan

[cyclebrain]

I spent a lot of time on this subject before I got my DAC working and
had questions about my method of calculation because my answers seemed
to low. After building and using I had to use a LOT more volume
control to listen. Measuring the output with a 0dB digital input I
only got 380mv out. I then repeated the same test using the analog
output from the built in DAC from my CD player to verify my test set
up and got the expected 2v output.
I still having problems with how to calculate this. The output is
rated at 7.8ma peak to peak. The 2 volt output is an RMS rating. So
one output side has a max current of 3.9ma and the other output is at
-3.9ma. The voltage across the I/V resistor is 4ma peak x 25ohm= 0.1v
peak. 0.1v peak x 6.67gain = 0.667v peak. 0.667v peak x .707 = 0.472v
rms. I measured only 0.380v rms though. After making my changes, I/V
resistor 50 ohms and amplifier feedback resistor 2.7K ohms I still
measured short. 1.75v rms with a 0dB digital input.
Anybody else have any input?

[cyclebrain]

I got the damaged regulator replaced and changed the I/V resistor from
25ohms to 50ohms and changed the feedback resistors from 1Kohms to
2.7Kohms.
Still my measured output using a 0dB 1KHz digital input came up short
of my calculations and measured only 1.738V rms into a 47K ohm load
instead of the 2v spec. Still a lot better than the previous output of
384mV out. I decieded to continue with my testing even though the
output level was still low and would thus effect the S/N ratio.
If I found any other problems while testing I would like to take care
of all of them at the same time.
At this time I would like to state that my test data provided is for
reference only and to identify any obvious problems.
I used an HP 3478A true RMS digital multimeter for the 0 to -30dB
linearity tests and for the frequency response tests at -30dB. This
voltmeter is rated from 20Hz to 100KHz. Below -30dB the noise level of
the voltmeter began to swamp the signal and made it useless. I also
used my computer running RTA spectrum analyzer and oscilliscope
software. This program worked surprisingly well despite using a cheap
computer sound card.
Now for the data.
Channel balance error - 0.08dB
D.C. offset Left 0.4mV, Right 1.3 mV
Amplitude Linearity 0dB to -10dB in 1 dB steps Left -0.01 +0.005
dB Right -0.002 +0.005dB
-10dB to -30dB in 10 dB steps Left -0.5
+0 dB Right -0.43 +0.02dB
-30dB to -90dB in 10 dB steps (used
spectrum analyzer) looked good, no values.
Frequency Response at -30dB level 10Hz to 20KHz +/- 0.03 dB

While doing harmonic and I.M. distortion tests I got bored with trying
to quantify values with a piece of equipment that is not up to par and
just started looking for things that didn't look right.
Harmonic levels looked very good, the only harmonic appearing being
the 3rd but well down.
Doing the 19khz+20khz I.M. test I was only able to view the difference
value, 18khz. It was well down also.
I had no digital test signals above 20khz to check for high pass
filter response.
60Hz and 120 Hz leakage looked well down into the noise floor also.
All in all I think a good peice of equipment.
Well done Evan.

[Evan Zamir]

John, so now I'm very curious. Have you done these types of measurements on other DAC's. It would be interesting to know how it compares. Anyway, that' s probably not the most important thing. I guess as long as you like the sound, I'm happy :)

As for the output, I had done my calculations mostly using TINA (the free spice program from TI). I think based on your results, I will recommend the I/V resistor be 50R and the feedback resistor 3K. This, of course, if for PCM1794. PCM1798 has half the current output of 1794, so the feedback resistor would need to be doubled again. Like I said in an earlier post (maybe), the output of the original DAC I built with PCM1798 was well enough to drive my amp only using passive attenuators. And that was with a 25R I/V resistor and 1.5K feedback resistor.

Thanks, so much for doing these measurements, and taking the time to work with the DAC. I appreciate it, and I'm sure everyone else working on the DAC does, too.
-evan

+0 dB            Right  - 0.43   +0.02dB

[cyclebrain]

No, I have not done any previous measurements on a DAC. I just wanted
to be sure that this one wasn't doing anything out of the ordinary and
took me a while to figure out how to provide the digital test signals.
Most of my previous testing has been room response measurements. What
a pain/art/science/comprimise effort that is. I don't completely trust
my ears and need some basic measurements to back up what I hear, but
am not anal about the data. My day job is "senior system integration
and test engineer" for a major defense contractor.
Aircraft based synthetic aperture radar. Sometimes my, that's close
enough, basic rule of thumb. troubleshooting gets me in trouble, but
it usually is more efficient when problem solving.
Of course as with everything there are tradeoffs.
A current source likes a 0V load for linearity and for a good input to
the output amp we would like a large input voltage. Two contradictory
situations.
Now that I am comfortable that the DAC isn't doing anything bad
(actually performs very well) I am now doing my subjective testing and
am quite pleased.
I'm glad to have contributed to this project. Your work creating the
PC board was a fantastic find for me and I really appreciate it.